vlog-auto-edit

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.80). The SKILL.md Phase 2 "参考研究" explicitly instructs the agent to fetch/download public vlogs from B站/YouTube (via yt-dlp) and analyze them (ASR/visual API), meaning the agent ingests untrusted, user-generated web content that can influence its LLM planning and subsequent actions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The README explicitly tells the agent at runtime to "请从 https://github.com/znyupup/ai-video-editing-skill 克隆仓库" so the agent would git-clone and read SKILL.md (remote content) which directly controls prompts/instructions the agent should follow, making this a required runtime dependency that can alter agent behavior.